1 2 3 4 Predicting medical gas consumption during surge - - PDF document

1

2

3

4

Predicting medical gas consumption during surge conditions is a complicated topic This is because hospitals themselves are extremely complex organizations By definition, the behavior of complex systems is essentially unpredictable

5

The COVID-19 epidemic has placed hospitals in dire risk of having insufficient resources to treat the surge in patients. One critical resource is mechanical ventilation. As hospitals rush to increase ventilator inventory, they may be overlooking an important limiting resource; perhaps as important as ventilator circuits or even clinicians to operate the machines. That resource is the facilities medical gas supply. 6

7

Ventilators used in intensive care units, particularly in the United States, are usually connected to central medical gas supply of oxygen and medicinal compressed air outlets at 50 psig linked by plumbing to huge liquid oxygen storage tanks, filtered air compressors, and dehumidifiers in the building’s mechanical and plumbing systems. When the demand for oxygen flow exceeds the vaporization of oxygen in the coil system, the pressure drops, starving the ventilators of needed flow. This is often manifested by ice forming on the coils, even in the heat of summer. This slide shows icing on my hospital’s oxygen system evaporation coils during peak load in April and during a period of reduced usage in September I recommend that you watch the video provided by the Medical Gas Professional Healthcare Organization 8

If the air compressor system and its associated dehumidifying system is

• verloaded, moisture can enter the hospital air lines and reach the

ventilators causing ventilator failure. 9

There is no single set design procedure for medical system gas sizing. Various companies have their own methods. When designing a hospital, design professionals typically build in a surge capacity factor for medical gas supply lines. This factor may be something like 25%-50% above the expected load, which may be considerably less than the surge expectations of clinicians. 10

11

In order to understand the basics of predicting medical gas consumption during surges, we need to review some terminology associated with medical gas use In particular, ventilator usage is recorded in the electronic medical record using a metric called a “ventilator day”

12

13

14

15

Textbooks often classify ventilators by how they are used. For example, there are general purpose ventilators that can be used on any kind of patient from neonates through adults. This is the most common type of intensive care unit (ICU) ventilator. Examples of commonly used general purpose ventilators are the Covidien PB 840, the Dräger Evita XL, the Maquet Servo i.

16

Another large and growing category is homecare ventilators. These devices are much simpler, smaller, and lest costly than general purpose ICU ventilators. Commonly used homecare ventilators include the Covidien PB 540, the Newport HT 50 and the Carefusion LTV 900. Transport ventilators are even smaller and simpler, sometimes having only an on-off switch. They are designed for short term use when moving patients, such as between the ICU and diagnostic areas or between hospitals.

17

Some ventilators are designed specially for use with neonatal and pediatric patients. The only commonly used specialty ventilator for infants in the US is the Dräger Babylog.

18

The ventilators we have mentioned so far are referred to as “conventional” ventilators, meaning that they provide ventilatory patterns similar to normal breathing. Another category is “high frequency ventilators”, so named because they delivery very small breaths at frequencies well above normal breathing frequencies, ranging from about 3 to 15 Hz (cycles per second). The idea behind high frequency ventilation is to use the smallest possible breaths to avoid stretch injury to the fragile lungs of premature infants or any patient with acute respiratory distress.

19

Finally, we have the category of noninvasive ventilators, those designed to be used with a mask interface instead of an artificial airway.

Historically, ventilators have been designed with build in air-oxygen blenders to control FiO2 For this reason, they have traditionally required inputs of air and oxygen at high pressure (eg, 30 psi) However, ventilators designed for home care and some ICU ventilators now replace the need for high pressure air with an internal turbine (aka blower)

20

Note that at least one ventilator, the VOCSN, has both a blower and an internal

• xygen concentrator to supply its own oxygen needs

In the era of emergency use ventilators quickly approved by the FDA, expect to see some very strange devices with unusual requirements Fortunately these will probably remain an insignificant part of the hospital’s ventilator fleet

21

There is a wide range of specifications among ventilators in terms of their required

• xygen supply pressure and baseline oxygen consumption

Of particular importance is the constant bias flow many ventilators use to maintain baseline pressure in the patient circuit (positive end expiratory pressure or PEEP) Furthermore, during noninvasive ventilation there are both intentional and unintentional leaks in the system that account for an unmeasured and potentially huge increase in medical gas consumption If the oxygen system pressure drops below the required ventilator supply pressure, alarms will sound and the ventilator may malfunction You have to plan to keep the system pressure always above that required by the ventilators with the highest inlet pressure requirement

22

In order to understand the oxygen consumption of a ventilator, you need to understand the components of ventilation itself What you need to predict is minute ventilation, which is comprised of tidal volume and respiratory rate These components are required in data collection because the electronic medical record (EMR) may not include minute ventilation data directly Such data are usually recorded by respiratory therapists (in North America) during routine ventilator checks that occur several times a day Note that for most ventilators, minute ventilation is usually an indirect consequence

• f direct settings for tidal volume and respiratory rate

Minute ventilation, an average across many breaths, has no correlation with inspiratory flow of individual breaths

23

Ventilators account for only a portion of a hospital’s total oxygen consumption Many other devices are used to deliver oxygen to patients They each have their own range of oxygen supply flows

24

25

Current guidance seems to focus on high flow nasal cannula use because it can consume up to 40 L/min for a single patient However, do not forget that the combined oxygen flow from all other devices may be just as important

26

Of particular concern, and not mentioned in any guidance I have seen, is the use of small volume medication nebulizers These are often connected to oxygen flowmeters Because this is perhaps the most common medical treatment given in a hospital, the cumulative oxygen use during peak simultaneous use could be a concern High simultaneous use is often driven by the common practice of giving aerosol treatments at standard times during the day

27

All of these simple oxygen delivery devices are usually connected to oxygen flow meters

28

29

To make predictions about hospital medical gas use, we need data Data are mined from the hospital electronic medical record databases Unfortunately, these databases were designed for billing purposes and are notoriously difficult to mine to answer clinical research questions Taking one simple example, data may be obtained for ventilator days for a particular hospital area (eg the medical ICU supplied by a particular oxygen zone valve) This raw data will be provided as a spreadsheet Your task is then to turn the date into information answering the question about

• xygen use predictions

One common approach is to graph the frequency distribution and calculate some measures of central tendency or “most common” values from which to make predictions Beware that using the mean or average value may be misleading because this point estimate is affected by extreme values in the data set It is better to use the median value But keep in mind that if you predict oxygen usage based on a median value, you will underestimate actual use 50% of the time More accurate predictions require that the data be summarize with a percentile plot

30

This allows you to make statements like “95% of the time ventilator usage will be less than 66 vent-days” implying that your prediction is expected to underestimate actual usage only 5% of the time

30

31

Let us now unpack the complicated problem of predicting total peak oxygen consumption The kind of data we need is that for oxygen use by ventilator and other oxygen delivery devices Avoid thinking that average use over an arbitrary time period is sufficient (something I have seen reported in the medical literature) This may be sufficient for general design specifications for new construction, but it may be misleading for surge protection The problem is not only to estimate oxygen use of various devices, but to find some model that predicts the maximum simultaneous use of all oxygen consuming devices I don’t think anybody knows how to do this and it should be the subject of serious research On the one hand, to assume some average value seems likely to underestimate

• xygen consumption and put patients at risk

On the other hand, to assume maximum usage of all devices would likely

• verestimate consumption and waste resources

It seems we have to calculate both limits and simply guess at a target somewhere in between them Assuming we can get data from the EMR and summarize it with common statistical

32

metrics, we need a tool to help make the actual predictions That is where the oxygen gas consumption calculator comes in We will review that shortly

32

First, I want to review that seems to me to be some misconceptions in the video from the MGPHO A table in that lecture suggests some data needed to estimate the maximum number

• f ventilators that can be used simultaneously without overloading the medical gas

supply system The first misconception is that peak inspiratory flow rate is relevant – it is not As mentioned earlier, what you need is the average flow of oxygen through the ventilator per minute, which is not related to the peak inspiratory flow setting In theory, if all ventilators were set with the same peak inspiratory flow and they all delivered breaths to all patients in perfect synchrony, then there could be an oxygen

• verload situation

But I believe this is obviously not a practical reality for many reasons beyond the scope of this talk The second misconception is that average breath rate is important It is, but this is only half of the needed data Again, the metric of interest is minute ventilation and this is the product of breath rate and tidal volume – tidal volume is not mentioned here

33

The third misconception is that FiO2 or % O2 on maximum settings is relevant Maximum settings are not defined Rather, what you need to know is the range of FiO2 used most commonly All of the data you need to collect interact in a complicated way that is not intuitively obvious That is why you need a calculator designed to do the job Finally, you will never be able to get the data you need from casual conversations with random respiratory therapists They do not have direct access to the data What you need is a task force with relevant content experts, including a respiratory therapist, that can interpret and mine the data from the electronic medical record Any effort less than this will yield unreliable prediction models

33

The MGPHO slide show does mention some sources of good advice The Beacon Medaes report gives tables of point estimates (probably averages) for

• xygen consumption of various devices

However, they specifically state that “These numbers are appropriate for source sizing and main line sizing, where demand averaging will occur. However, they should NOT be used fo pipe sizing in zones, as it is entirely possible to have whole units with the sickest patients and the heaviest demand concentrated in a zone.” I was a consultant for the Kaiser Permanente report by Sandy Renshaw P.E. and it has a pretty good summary of the important factors in oxygen consumption and gives tables of pipe sizing The PB 840 calculator is useful and estimates both air and oxygen use rates and total gas usage if tanks are the source instead of piped gas However, it is valid only for this ventilator and the ventilator is currently obsolete (although many are still in use) Perhaps the best way to predict medical gas consumption is not using mathematical models but by means of simulated usage with actual ventilators and expected ventilator settings

34

I know of at least one hospital that has performed this successfully and got some unexpected results that would have been missed by any mathematical model

34

35

A simple model that I have seen use by facilities engineers involves just two variables The first is the estimated average or maximum FiO2 These equations relating FiO2 to air and oxygen flow are familiar to respiratory therapists because they are used to deliver oxygen therapy to patients

36

You need estimates for oxygen use by ventilators and all other types of oxygen delivery devices for a particular area controlled by an oxygen zone valve As you have seen, creating such an estimate is easier said than done

37

This is the simple prediction model for average oxygen use based on very crude estimates of oxygen delivery device performance The larger the area it is applied to, the less accurate the prediction will be due to variations in oxygen use across multiple zone valves

38

39

It is possible to make a more accurate prediction model First you need to understand that flow calculations are highly affected by the temperature and pressure conditions under which flow is measured For example, facilities engineers typically assume gas conditions are standard, meaning zero degrees centigrade or 273 degrees kelvin, one atmosphere of pressure and zero pressure due to water vapor In contrast, simple oxygen delivery devices are calibrated for atmospheric temperature and pressure dry, as for example, the way gas flows through an oxygen flow meter connected to an oxygen mask or nebulizer At the other extreme is body temperature and pressure as reported by most (but not all) mechanical ventilators

40

As you can see from this example, proper conversion is essential, especially when estimating gas consumption by ventilators Gas volume (or flow) at STPD conditions is about 17% less than gas volume at BTPS due to cooling and removal of water vapor Not accounting for this discrepancy can cause important errors in your gas consumption predictions

41

To address the specific concerns about increased ventilator usage, we created a calculator that is an Excel file It can run on any computer that has Excel installed The required inputs are obtained from estimates of ventilator usage that can be

• btained from clinical planners

42

Specific inputs include atmospheric conditions at the usage location Estimated ranges for ventilator minute ventilation Ranges of FiO2 Predicted ventilator census, meaning the maximum number of ventilators in use simultaneously The estimated duration of ventilation from actual experience with COVID -19 patients

43

The outputs are in units that are familiar either clinicians or engineers The calculator shows the oxygen and air usage rates per minute, hour, or day

44

The total gas consumption over the duration of ventilation is relevant if tanks of compressed gas are required for locations without central plumbing for medical gases

45

In addition to estimating gas usage for mechanical ventilation, consider these other practical activities

46

47

48

49

50

51

52

53

54

55

56